Four Ways to Make Money in the Agentic Economy

TLDR: The agentic economy breaks down into four quadrants based on two axes: digital vs. physical work, and renting vs. owning. In three of those four quadrants, the compounding upside quietly flows upstream to whoever owns the underlying asset. Only one position, the founder who owns proprietary data and distribution, reliably lets you keep what you build.

I spend a lot of time thinking about where software value actually accumulates, and this framework from TheCircuit's MetaCircuits cuts through a lot of the noise I hear at conferences and on podcasts. The framing is simple: cross "bits vs. atoms" with "rent vs. own" and you get four distinct positions in the agentic economy. What's uncomfortable about this framework is what it reveals about most people's current bets.

Start with the digital rented quadrant, which is where most of us live right now. ChatGPT, Claude, Cursor, Perplexity. You pay monthly, you point it at your work, you go faster. The income play is leverage, more output per hour, higher quality, same you. But here's the catch: the only thing that sticks is your accumulated context, your prompts, your workflows, your institutional knowledge baked into those conversations. The AI lab owns the model, owns the infrastructure, and owns the data flywheel. You're renting productivity. And as it turns out, about 47% of individual AI tool users convert to enterprise contracts, nearly double the SaaS conversion rate. You're not the end customer in that story. You're the top of the funnel.

The founder's quadrant is digital owned agents, and this is where I think the most interesting individual bets are being placed right now. The shift YC keeps pushing is from selling tools to selling work. A clinic doesn't want to buy software that helps schedule appointments. They want the appointments scheduled. Office software runs maybe $500 a year per seat. A receptionist costs $47,000 a year in North America. That delta is the whole game. The agentic moat in this space is proprietary data, workflow integration, and distribution. If you own all three of those, the upside is genuinely yours. This quadrant is already at roughly $45-50 billion in revenue in 2026, and it's the fastest growing of the four.

Physical agents introduce a different set of dynamics. Renting physical autonomy looks great on an operational level. Robots-as-a-service, equipment-as-a-service, power-by-the-hour. Rolls-Royce perfected this with jet engines decades ago. Airlines pay for hours of thrust, Rolls-Royce keeps the engine, the maintenance contracts, and critically, the data. The operator position here can work if you own the scarce complement: the route, the permit, the local relationships, the proprietary operational data. Without that edge, you're just taking on financial risk while someone else holds the compounding asset. And the history of vertical integration in tech should make anyone in this position nervous. Amazon mined seller data to launch private-label products. Apple Sherlocked third-party apps. Netflix went from distributor to rival studio. The entity whose machine you're leasing has every incentive to eventually compete with you directly.

The capital-intensive physical owned quadrant is where the deepest moats sit, and it's largely not a place individuals can play. Anduril, Symbotic, Waymo at a $126 billion valuation, Amazon's warehouse robot fleet. These stack two compounding barriers simultaneously: scarce, capital-intensive physical assets and a data flywheel that grows more valuable with every hour of operation. The best play in the entire framework is to own the asset and rent it out as a service simultaneously. Waymo owns the cars and sells rides. AWS owns the servers and sells compute. They collect from every operator on the platform while the asset appreciates and the data compounds.

What I find genuinely interesting is the observation about distribution changing. Selling to agents is not like selling to humans. You can't win on brand, on impulse, on a clever ad. An agent buying on your behalf is going to evaluate on machine-readable signals: price, reliability, reputation, API availability. The discovery layer that emerges from agent-to-agent commerce isn't plumbing. It's the new global marketplace, and whoever owns identity, trust, and payments in that layer is playing for enormous stakes.

Key takeaways:

In three of the four agentic economy quadrants, durable compounding value flows upstream to whoever owns the underlying asset, not the operator or renter
The founder building digital agents on proprietary data and distribution is the one position where individuals can genuinely capture long-term upside
AI venture funding reached 80% of all global VC in Q1 2026 ($242 billion), and the target is the $60-65 trillion global wage bill, not the sub-$1 trillion software market
Distribution fundamentally changes when buyers are agents: brand and impulse don't matter, machine-readable signals like price, reliability, and API access do

Why do I care: As a senior developer and architect, I spend a lot of time evaluating where to put my energy, and this framework is genuinely clarifying. If I'm building something with AI, the question isn't "what model do I use" or "how good is my agent." The models commoditize. The question is whether I own the data, the workflow integration, and the distribution channel. That's the checklist I'm running against every new project now. The alternative, building on someone else's platform without owning the scarce complement, is just taking on operational risk while the platform owner keeps the upside. I've seen this movie before in web development, and I'd rather write the next one.

There are only four ways to make money in the agentic economy